165

Psychology (Fach) / Research Methods (Lektion)

Diese Lektion wurde von Julia2308 erstellt.

Lektion lernen

zurück  |  weiter 3 / 4

• Nonreactive measurements - Physical traces e.g wear and tear on seperate sections of an encyclopedia > observations that are not visible for the person being observed
• Nonreactive measurements - Simple unobstrusive observation - observations made that were inconspicous to the participants and there was nothing contrived or manipulated in the situation - “Contrived unobtrusive observations” are unobtrusive measures in manipulated situations
• Judgement studies > studies in which judges are used to evaluate or categorize the variables of interest > positive correlation between bias toward a category and accuracy in that category > (observed frequency - expected frequency)2 _________________________________         = judge's accuracy     expected frequency > the greater the score, the greater the bias
• Rating biases - Errors of Liniency > Rating of someone familiar as more positive than should be > opposite of ''strictness error'' > can be overcome: > by using numerical and graphical scales > ask judges to rate positive and negative qualities on two seperate unipolar scales
• Rating Bias - Halo Effect > refers to a type of response set in which the person being evaluated is judged in terms of a general impression > can be overcome by: > forced choice format
• Rating Bias - Severity Error > being aware of the halo effect, judges may ''lean over backward'' and be overly negative (''strictness bias'')  
• Rating Bias - Errors of central tendency > occurs when judhes hesitate to give extreme ratings and instead tend in the direction of the mean of the total group > can be overcome by: > stretching the positive end of a rating scale  
• Rating Bias - Logical error in rating > refers to the problem that many raters are likely to give similar ratings to variables or tratis that seem logically related in their own minds > can be overcome by:  > construct precise defintions and make sure that the instructions are also clear, precise and explicit > difference between log.e.i.r and halo effect: halo effect results from subjective feelings, whereas the other from conscious evaluation of relatedness
• Forces choice judgements > requires judges to make judgements between two or more equally favourable characteristics > developed to overcome halo effect
• Dimensional (judgement) rating scales - Numerical scales > judges or raters work with a sequences of defined numbers > has numbers as anchors and a following description which is used to give those numbers meaning and a specific context to evaluate
• Anzeige
• Dimensional (judgement) rating scales - Magnitude > a concept in which the upper range of the score is not defined but left for interpretation of the judge (open-ended) > statements are evaluated in porportion to the magnitude of other statements  
• Dimensional (judgement) rating scales - Graphic formats > simply straight lines (mostly invisibly consisting of different areas) in which the judge or the participant should indicate his/her position (attitude) with regard to construct in question (anchor) > Unipolar scales: contain gradual distributions of one aspect > Bipolar scales: ranges from a great amount of a characteristic to a great amount of its opposite (cold vs. warm)  
• Composite Variables Composite measures of variables are created by combining two or more separate empirical indicators into a single measure. Composite measures measure complex concepts more adequately than single indicators, extend the range of scores available and are more efficient at handling multiple items. > composite measures of variables are created by combining two or more seperate empiricial indicators into a single measure > to form composite variables, you start standardizing the scores of all available variables and then replace these score with the overall mean > + : measure compley concepts more adequately than single indicators, extend the range of scroes availabe and are more efficient in handling multiple items > +: conceptually - if variables cannot be discriminated from one another > +: practically - you are able to obtain more accurate estimates of the relationships with other composite variables and you reduce the number of predictors
• Composite variables > Intra/Intermatrix > Intracomposite average means the average between variables within a single composite > Intercomposite average characterizes the level of relationship between one composite variable and another > only when individual varibales are perfectly correlated with each other is there no benefit of forming composite variables > in general: the larger the number of seperate variables that are combined into a composite, the greater the increase in effect size r(composite) obtained will be  
• Composite variables - Quantifying the clarity of composites - the r method: the point-biseral correlation is computed > rpb is computed between the mean correlations of the intra/Intermatrix (continuous) with their dichotomously located position on the principal diagonal, versus off the diagonal > he more positive the correlation, the higher the “intra”(on the diagonal) than the “inter” (those on the off diagonal) - the g method:  g = r(inta) - r(inter) / S(combined) > indicates that the average relationship of variables within a composite to one another is g standard deviations larger than the average relationship of variables between composites to one another - range-to-midrange ratio > can be used when both r and g method can not be used, hence the S is missing > r(intra)- r(inter) / ((r (intra) + r (inter)) / 2) > a particular raw difference between the r(intra) and r(inter) is relatively more meaningful when the difference between the mean of these rs is smaller rather than larger  
• Experimentation > a systematic study design to examine the consequences of deliberately varying a potential cause agent > Features: Variation in the treatment Posttreatment measures of outcomes Inferential techniques are common features Atleast one unit on which observation is made
• Randomized experimental designs Between subjects design > regarded as the “gold standard”, such as randomly assigning the subjects to a specific treatment group or “placebo > most common randomized experimental design > use of placebos is controversial  
• Randomized experimental design Wait-list control groups > another possibility; combined with repeated measurements, these allow us to gain further information about the temporal effects of the drug > delaying the treatment for one group by giving them a placebo first, but the real drug later on after the outcomes have been measured
• Randomized experimental designs Within-subject design (or crossed design) > used when each participant is receiving more than one treatment or is in more than one condition (e.g. being experimenter and subject at the same time) > problem: the order in which the conditions are administered may be confounded by the condition effect > Order (carryover) Effects:  - Practice effects - Fatigue effects - Assimilation and contrast effects - Anticipation effects - “Catching on” effects > solution: counterbalancing > rotating the sequence of the conditions in what is called a latin square > every order possible ins administered in different sequences the subjects are randomly assigned to >
• Randomized experimental design Factorial design > includes more than one variable (or factor) which is measured by including more than one level of each factor (e.g. 2x2). Also known as “full factorials” > Fractional factorial designs only use some combinations of factor levels > Mixed factorial designs, consisting of both between- and within-subjects factors.
• Randomization > gives each treatment an equal chance of being allotted to any subject > a way of dealing with unsuspected sources of bias, although it cannot guarantee that all important natural differences among subjects will be exactly balanced out > variation on the randomized design: to use pretest measurements to establish baseline scores for all subjects  
• Experimental control - 4 types - control conditions - original meaning: ''to check'' - in research control means to keep something constant - behavioural control - refers to the shaping of learned behaviour based on a particular schedule of reinforcement designed to elicit the behaviour in question - control also refers to a standard of comparison
• Experimental control - 4 types - control conditions - original meaning: ''to check'' - in research control means to keep something constant - behavioural control - refers to the shaping of learned behaviour based on a particular schedule of reinforcement designed to elicit the behaviour in question - control also refers to a standard of comparison
• Mill's method of agreement and difference > refers to causal relationships - Method of agreement: if X, then Y - if we find two or more instances in which Y occurs, and if only X is present on each occasion, X is implied as a sufficient condition on Y  = experimental condition - Method of difference: If not X, then not-Y - if the presumed effect (Y) does not occur when X is absent, then we suspect X is not just a sufficient condition, but a necessary condition for Y to occur = control group > collectively referred to as joint method of agreement  
• Mill's method of agreement and difference > refers to causal relationships - Method of agreement: if X, then Y - if we find two or more instances in which Y occurs, and if only X is present on each occasion, X is implied as a sufficient condition on Y  - Method of difference: If not X, then not-Y - if the presumed effect (Y) does not occur when X is absent, then we suspect X is not just a sufficient condition, but a necessary condition for Y to occur  
• Moderator Variables > may alter the effect size correlations > alters the relationship between cause-and-effekt variables (these are variable)
• Mediater variables > are defined as factors that intervene between the independent variable and the outcome variable in a causal chain (there are systematic)
• Threats to External Validity (Generalization) - External validity: validity of inference about whether the cause-effect relationship holds over variation in persons, settings, treatment variables and measurement variables > often used as a synonym for generalization - Statistical generalizability / Conceptual replicability / Realism: > mundane realism: the extent to which an experimental treatment is apt to be encounteres in an analogous form in a natural setting > + experimental realism: refers to the psychological impact of the experimental manipulation on the participants > if we want to generalize the results of a psychological experiment to a population of subjects and a population of stimuli, then we must sample from both populations - Representative research design describes an idealized experimental model in which everything is perfect - Ecologically valid are experiments that satisfy this criterion of representativeness  
• Threats to Internal Validity - threats to internal validity compromise our confidence in saying that a relationship exists between the independent and dependent variables - threats to internal valitidy imply faulty cause-and-effect assumptions between independent and depedent variables - high internal validity implies strong evidence for a cause-and-effect relationship between variables 1) History: plausible source of error due to an uncontrollable event (between pretest and posttest) 2) Maturation: refers to intrinsic changes in the subjects, such as growing older, stronger or more experienced (between pretest and posttest) 3) Intrumentation: changes in the experimental instruments (between pretest and posttest) 4) Selection: unsuspected differences between participants in each condition 5) Regression toward the mean: : In an experiment involving reading instruction, subjects grouped because of poor pre-test reading scores show considerably greater gain than do the groups who scored average and high on the pre-test (example) > only affects one group designs ( one shot case study/ one group pretest/posttest w/o control group) > Solomon's extended control group design controls for threats to internal validity by extending the two-group design to three-group design and four group design
• Threats towards Internal Validity (added in Lecture) - Experimental mortality: participants fail to complete study - Demand characteristics: cues in the design tell participants how to behave -  Experimenter Bias: Experimenter subtley influences outcome
• Anzeige
• Artifacts - refers to research finding that result from factors other than the ones intended by the researchers, usually factors that are quite extraneous to the purpose of the inverstigation - factors that are evenly distributed across conditions and result in inaccurate findings > Hawthorne effect: subjects behave differently because they know that they are subjects of an inverstigation - Three distinct sources of artifacts: (Saul Rozenzweig) > observational attitude of the experimenter > motivational attitude: people may try to outguess the inverstigator, to figure out how their behavious will be evaluated > errors of personality influence: investigators sex and race
• Threats to Internal Validity - Demand characteristics - subtle, uncontrollable tast-orienting cues in an experimental situation (could produce artifacts) - good subject effect: subjects often go through remarkable lengths to comply with demand characteristics - strategies to control for demand characteristics/good subject effects: > quasi-control strategy: some subjects act as co-investigators = talk to other subjects after study, to find out whether ''good subject effect'' was a motivator behind the reactions to the study > preinquiry: some prospective subjects are seperated from the subject pool; research procedure and purpose are described and they are asked how they would react/ would be likely to behave in the experiment > sacrifice group of quasi controls: people pulled out at different points of time in the experiemtn and asked abou thteir perceptions of the experiment up to that point - volunteer subjects are the more compliant participants - evaluations apprehension: effects resulting from the participants anxieties about how they would be evaluated  > solution: ensure confidentiality    
• Threats to Internal Validity - Interaction experimenter effects - artifacts attributable to the interaction between experiments and their subjects 1) biosocial attributes: gender, race, age 2) psychosocial attributes: personality & temperaments 3) situational effects 4) modelling effect: it sometimes happens that before the experimenters conduct their studies, they try out the tasks themselves - if a modeling effect occurs, it is most likely to be patterned on the researcher opinion 5) experimenter expectancy effect: the experimenter's expectations may come to serve as a self-fulfilling prophecy
• Threats to Internal Validity - Experimenter expectancy effects - reduction of experiments expectancy effects: > increasing the number of experimenters > monitoring the behaviour of experiments > analyzing the experimenters for order effects  > maintaining blind contact:  single blind study /double blind study > minimizing experimenter-participant contact > employing expenctancy control groups: the most basic is to use a 2x2 design in which subjects are divided by treatment condition and expectancy condition (by the experimenter)  
• Nonrandomized research - Quasi -Experiment > refers experiments that are lacking the full control over the scheduling of experimental stimuli that make randomized experiments possible > ''assosciation'' implies covariation to some degree > ''methodological pluralism'': the assumption that because all research designs are limited in some ways, we need to draw on multiple methods for empirical investigation
• Non-randomized research - Nonequivalent groups design > because randomization is not always possible, non-equivalent groups design is an alternative: - resembles between-groups design except that researcher has no control over how the subjects are assigned to different groups - INCOMPLETE
• Non-randomized research - Nonequivalent groups design > because randomization is not always possible, non-equivalent groups design is an alternative: - resembles between-groups design except that researcher has no control over how the subjects are assigned to different groups - INCOMPLETE
• Non-randomization research - Interrupted time series - makes use of sampled measurements obtained at different time intervals before and after the intervention - 4 important considerations: > sufficient number of time points and of observations in order to use a particular data analytic method > the same units hace to be sampled thoughout the analysis to ensure that the observations and the time points are equally spaced > the time points have to be sensitive to the particular effect being studies > the measurements should not fluctuate very much as a results of instrumentation changes - 3 steps: > define period of observations > collect data > analyze with model is the most efficient in explaining the results - assumption: a series of observations must be stationary - value of the observations fluctuate normally about the mean > secular trend: most time series show systematic (non-stationary) increase and decreases in the level of the series > secular trend can be made stationary by differencing - consists of subtracting the first observation from the second observation, the second from the third, and so forth.. - autocorrelation: refers to whether the data points or observations are dependent on one another (autocorrelated) or instead can be assumed to be independent > regular: describes the dependency of adjacent data points or observations on one another > seasonal: describes the dependency of observations or data points seperated by one period or cycle
• Non-randomization research - Interrupted time series - makes use of sampled measurements obtained at different time intervals before and after the intervention - 4 important considerations: > sufficient number of time points and of observations in order to use a particular data analytic method > the same units hace to be sampled thoughout the analysis to ensure that the observations and the time points are equally spaced > the time points have to be sensitive to the particular effect being studies > the measurements should not fluctuate very much as a results of instrumentation changes - 3 steps: > define period of observations > collect data > analyze with model is the most efficient in explaining the results - assumption: a series of observations must be stationary - value of the observations fluctuate normally about the mean > secular trend: most time series show systematic (non-stationary) increase and decreases in the level of the series > secular trend can be made stationary by differencing - consists of subtracting the first observation from the second observation, the second from the third, and so forth.. - autocorrelation: refers to whether the data points or observations are dependent on one another (autocorrelated) or instead can be assumed to be independent > regular: describes the dependency of adjacent data points or observations on one another > seasonal: describes the dependency of observations or data points seperated by one period or cycle
• Non-randomization research - Single case experimental design - individual subject is first assessed under certain baseline conditions against which any other subsequent changes in behaviour can be evaluated after an environmental treatment is manipulated - reversal design: A-B-A design- A= baseline B= treatment > the baseline pattern is first identified - a sufficient number of data points is required to establish the pattern of negative and positive behaviour > statistical averaging of people washes out individual differences > Advantages: Processual: effects are monitored during the entire phase of treatment and observation Cost-effective: in terms of the number of subjects needed to test functional hypothesis > Criticism: because single-case studies focus on a single or few organisms, the results might not be generalizable beyond the limited circumstances of the observation > replication is required: - Direct replication: repetition of the same study- direct replication on the same subjects (intrasubject replications) presumably identifies the reliability of functional relationships over time - Systematic replications: varying an aspect, or the scheduling of the stimulus conditions or treatments- systematic replications on other subjects would identify the generality of the functional relationships across variations of independent and dependent variables and over other subjects
• Non- randomization research - Cross-lagged correlational designs -INCOMPLETE DUE TO NO UNDERSTANDING - 
• Non-randomization research - Invisible variation - path analysis: correlation strategy used to infer causations from nonexperimental data - method: statistically remove some assosciations betwen the variables and then appraising the partial correlations > attempt to rule out alternative pathways of causal inference on the basis of quantative models> GOAL: settle on the plausible causal pathway that is most viable > third variable problem: invisible variables problem - some variable that is correlated with both A and B may be the cause of both > mediator variable: might explain a causal effect of some variable X on an outcome variable Y > It is not possible to prove that a causal hypothesis is true, but it might be possible to reject untenable hypotheses and, in this way, narrow down the number of plausible hypotheses
• Non-randomization research - Invisible variation - path analysis: correlation strategy used to infer causations from nonexperimental data - method: statistically remove some assosciations betwen the variables and then appraising the partial correlations > attempt to rule out alternative pathways of causal inference on the basis of quantative models> GOAL: settle on the plausible causal pathway that is most viable > third variable problem: invisible variables problem - some variable that is correlated with both A and B may be the cause of both > mediator variable: might explain a causal effect of some variable X on an outcome variable Y > It is not possible to prove that a causal hypothesis is true, but it might be possible to reject untenable hypotheses and, in this way, narrow down the number of plausible hypotheses
• Longitudinal Research Design - Longitudinal Research: has the purpose to examine people’s responses over an extended time span - Cohort: or “generation” is a group of people born about the same time and having had similar life experiences > Longitudinal data are usually collected “prospectively” but the data can also be gotten “retrospectively” from historical records > Longitudinal studies can be used to track several cohorts to see if generation differences are present  
• Longitudinal Research Design - Longitudinal Research: has the purpose to examine people’s responses over an extended time span - Cohort: or “generation” is a group of people born about the same time and having had similar life experiences > Longitudinal data are usually collected “prospectively” but the data can also be gotten “retrospectively” from historical records > Longitudinal studies can be used to track several cohorts to see if generation differences are present  
• Longitudinal Research Design - Longitudinal Research: has the purpose to examine people’s responses over an extended time span - Cohort: or “generation” is a group of people born about the same time and having had similar life experiences > Longitudinal data are usually collected “prospectively” but the data can also be gotten “retrospectively” from historical records > Longitudinal studies can be used to track several cohorts to see if generation differences are present  
• Cohort Studies & Longitudinal Research Designs - cohort tables: x-axis (age), y-axis (period) - fallacy of period centrism: threat to assume that the results of an analysis of one given time perios is generalizable to other periods 3 effects can be detected: > age effect: implies changes in average responses due to the natural aging process > time-of-measurement effect: implies some kind of impact pf events in chronological time (occuring at the time of measurement) > cohort effect: implies ''past history'' specific to a particular generation and contributes to all measurements of the generation - all three effects can not be measured simultaneously in longitudal designs: > Simple cross-sectional design: subjects at different ages are observed at the same time. (DIFFERENT AGES; SAME TIME) > Simple longitudinal design: subjects of the same cohort are observed over several periods (SAME COHORT; DIFFERENT TIME) > Cohort-sequential design: several cohorts are studied, the initial measurements being taken in successive years. (DIFFERENT COHORTS;DIFFERENT TIMES) iv. Time-sequential design: subjects at different ages are observed at different times. (DIFFERENT AGES; DIFFERENT TIMES) v. Cross-sequential designs: several different cohorts that are observed over several periods are initially measured in the same period (SEVERAL COHORTS; SAME PERIOD)  
• Subclassification on propensity scores - Propesensity scores: the procedure involves reducing all the variables on which the treated and untreated participants differ into a single composite variable that effectively summarizes all the differences on all the variables (or covariates) > the major limitation of the propensity score method is that it can adjust only for observed confounding covariates - Subclasses are formed: subclasses increase precision of the analysis, assuming that there is a reasonable number of subjects in each subclass and that the subclassifications of the different treatment groups overlap (= people in one subgroup are more alike on a particular variable) > 5 or more classes can reduce bias in raw comparisons by some 90%
• Probability sampling - Every sampling unit has a known nonzero probability of being selected - The units are randomly drawn - The probabilities are taken into account in making estimates from the sample  
• Paradox of Sampling > implies that the sample is of no use when it’s not representative of the population. In order to know that its representative, you need to know the whole population characteristics, but then you don’t need the sample in the first place

zurück  |  weiter 3 / 4